[ECOS] odd serial input bug in redboot 095308 20010607 on iPaq

[ECOS] odd serial input bug in redboot 095308 20010607 on iPaq

[David Bliss]

I've just installed RedBoot 09:53:08 Jun 7 2001 on my iPaq HPC.  I'm having
some trouble. 

When it's cold-started (battery switch turned on), it will
output its banner to the serial console:

+FLASH: 0x50000000 - 0x51000000, 64 blocks of 0x00040000 bytes each.
... Waiting for network card: ..........No network interfaces found

RedBoot(tm) bootstrap and debug environment - built 09:53:08, Jun  7 2001

Platform: Compaq iPAQ Pocket PC (StrongARM 1110)
Copyright (C) 2000, 2001, Red Hat, Inc.

RAM: 0x00000000-0x01fc0000, 0x0001ff00-0x01f70000 available
RedBoot>

But it will not accept any input from the serial console, only from the
keyboard.

If I boot into Parrot (where the serial interface works flawlessly) and then
boot RedBoot, it works.  If I warmboot after that (reset command) it works.
But every time I cold boot, no serial input.

Since running Parrot makes it work, I'm guessing that perhaps the UART 
initialization leaves something to be desired?

Anyway, it's a pain, though not a showstopper, but someone more knowledgeable
than I might want to take a look at it.

I'm more than happy to offer remote login to the machine if anyone wants
to look at its configuration.

Thanks,
David Bliss
University of Washington
Clinical Informatics Research Group
dbliss@u.washington.edu / david@dbsi.org

RE: [ECOS] odd serial input bug in redboot 095308 20010607 on iP

[Gary Thomas]

On 08-Jun-2001 David Bliss wrote:
> I've just installed RedBoot 09:53:08 Jun 7 2001 on my iPaq HPC.  I'm having
> some trouble. 
> 
> When it's cold-started (battery switch turned on), it will
> output its banner to the serial console:
> 
> +FLASH: 0x50000000 - 0x51000000, 64 blocks of 0x00040000 bytes each.
> ... Waiting for network card: ..........No network interfaces found
> 
> RedBoot(tm) bootstrap and debug environment - built 09:53:08, Jun  7 2001
> 
> Platform: Compaq iPAQ Pocket PC (StrongARM 1110)
> Copyright (C) 2000, 2001, Red Hat, Inc.
> 
> RAM: 0x00000000-0x01fc0000, 0x0001ff00-0x01f70000 available
> RedBoot>
> 
> But it will not accept any input from the serial console, only from the
> keyboard.
> 
> If I boot into Parrot (where the serial interface works flawlessly) and then
> boot RedBoot, it works.  If I warmboot after that (reset command) it works.
> But every time I cold boot, no serial input.
> 
> Since running Parrot makes it work, I'm guessing that perhaps the UART 
> initialization leaves something to be desired?
> 
> Anyway, it's a pain, though not a showstopper, but someone more knowledgeable
> than I might want to take a look at it.
> 
> I'm more than happy to offer remote login to the machine if anyone wants
> to look at its configuration.

I've never seen this and I _have_ put RedBoot on a number of iPAQ units.
Which is not to say, "I don't believe you".

What happens if you press the reset button (small indent on the bottom 
of the unit)?

The most likely cause is that when powered on cold, on your unit, the keypad
or the touch screen are presenting some events.  This will be interpreted by
RedBoot as an attempt to use the keypad/touchscreen for input and activity
on the serial port will cease.  Remember that RedBoot tries to use any available
I/O device when first booted and then "settles" for the first one that actually
gets any input.

As an experiment, you could try installing a version which subverts this
behaviour.  We could also try to learn what's coming in from where (keypad
or touch screen) to see what's up here.

Re: [ECOS] odd serial input bug in redboot 095308 20010607 on iP

[David Bliss]

On Fri, Jun 08, 2001 at 09:02:51AM -0600, Gary Thomas wrote:
> 
> I've never seen this and I _have_ put RedBoot on a number of iPAQ units.
> Which is not to say, "I don't believe you".
> 
> What happens if you press the reset button (small indent on the bottom 
> of the unit)?

Same behaviour as 'reset'.  If it was previously working, it continues to
work.  If it was previously not working it continues to not work.

> The most likely cause is that when powered on cold, on your unit, the keypad
> or the touch screen are presenting some events.  This will be interpreted by
> RedBoot as an attempt to use the keypad/touchscreen for input and activity
> on the serial port will cease.  Remember that RedBoot tries to use any available
> I/O device when first booted and then "settles" for the first one that actually
> gets any input.

Possible, but on first touchscreen input, the screen 'blinks' in the same
way that it did when settling on an input device (whether screen or touchpad)
when running from WinCE.

> As an experiment, you could try installing a version which subverts this
> behaviour.  We could also try to learn what's coming in from where (keypad
> or touch screen) to see what's up here.

I'd be glad to do so if you'll assist me in locating/preparing such an image
:)

Thanks,
david

Re: [ECOS] odd serial input bug in redboot 095308 20010607 on iP

[Gary Thomas]

On 08-Jun-2001 David Bliss wrote:
> On Fri, Jun 08, 2001 at 09:02:51AM -0600, Gary Thomas wrote:
>> 
>> I've never seen this and I _have_ put RedBoot on a number of iPAQ units.
>> Which is not to say, "I don't believe you".
>> 
>> What happens if you press the reset button (small indent on the bottom 
>> of the unit)?
> 
> Same behaviour as 'reset'.  If it was previously working, it continues to
> work.  If it was previously not working it continues to not work.
> 

That pretty much kills the speculation below.

Have you changed the baud rate, or is it still at the default of 38400?

Do you see any of the boot-up messages on the serial port when it freezes?  These
should be displayed, no matter what, up to the first prompt.

I do have a unit here where the serial is pretty squirrely, although not the
same as yours.  I've found that if I simply unplug/plug the serial (take it out
of the cradle, whatever), that it comes back fine.  Can you try this?

Do you have a network connection?  If so, you could try accessing the unit that
way when it is in this state and maybe get a glimpse into what it is doig.

What is the vintage of this unit?

>> The most likely cause is that when powered on cold, on your unit, the keypad
>> or the touch screen are presenting some events.  This will be interpreted by
>> RedBoot as an attempt to use the keypad/touchscreen for input and activity
>> on the serial port will cease.  Remember that RedBoot tries to use any available
>> I/O device when first booted and then "settles" for the first one that actually
>> gets any input.
> 
> Possible, but on first touchscreen input, the screen 'blinks' in the same
> way that it did when settling on an input device (whether screen or touchpad)
> when running from WinCE.
> 
>> As an experiment, you could try installing a version which subverts this
>> behaviour.  We could also try to learn what's coming in from where (keypad
>> or touch screen) to see what's up here.
> 
> I'd be glad to do so if you'll assist me in locating/preparing such an image
>:)
> 
> Thanks,
> david

Re: [ECOS] odd serial input bug in redboot 095308 20010607 on iP

[David Bliss]

On Fri, Jun 08, 2001 at 09:23:48AM -0600, Gary Thomas wrote:
> 
> Have you changed the baud rate, or is it still at the default of 38400?

Still at 38400.  Haven't tried anything else.  I've reset the baudrate via
fconfig several times, but always to 38.4.

> Do you see any of the boot-up messages on the serial port when it freezes?  These
> should be displayed, no matter what, up to the first prompt.

I see up through the first RedBoot> prompt, as pasted in the original mail,
and nothing beyond.

> I do have a unit here where the serial is pretty squirrely, although not the
> same as yours.  I've found that if I simply unplug/plug the serial (take it out
> of the cradle, whatever), that it comes back fine.  Can you try this?

One of the first things I've tried.  No help.

> Do you have a network connection?  If so, you could try accessing the unit that
> way when it is in this state and maybe get a glimpse into what it is doig.

Not yet.  I hope to next week, depending on when my PCMCIA sleeve arrives...


> What is the vintage of this unit?

Brand new.  S/N 4g15dw312091, on the off chance that means anything to you. :)
Purchasing happened through another department, but I believe it was ordered
directly from Compaq.

Interestingly, even when serial is working, I can't make it accept any
files via xmodem.  Gives 3 retries and then aborts back to the prompt
(at least, it's sitting at the prompt, very confused, when I abort the
upload on the PC).  

david

Re: [ECOS] odd serial input bug in redboot 095308 20010607 on iP

[Gary Thomas]

On 08-Jun-2001 David Bliss wrote:
> On Fri, Jun 08, 2001 at 09:23:48AM -0600, Gary Thomas wrote:
>> 
>> Have you changed the baud rate, or is it still at the default of 38400?
> 
> Still at 38400.  Haven't tried anything else.  I've reset the baudrate via
> fconfig several times, but always to 38.4.
> 
>> Do you see any of the boot-up messages on the serial port when it freezes?  These
>> should be displayed, no matter what, up to the first prompt.
> 
> I see up through the first RedBoot> prompt, as pasted in the original mail,
> and nothing beyond.
>
 
Sorry, that wasn't 100% clear from your first message.

>> I do have a unit here where the serial is pretty squirrely, although not the
>> same as yours.  I've found that if I simply unplug/plug the serial (take it out
>> of the cradle, whatever), that it comes back fine.  Can you try this?
> 
> One of the first things I've tried.  No help.
> 
>> Do you have a network connection?  If so, you could try accessing the unit that
>> way when it is in this state and maybe get a glimpse into what it is doig.
> 
> Not yet.  I hope to next week, depending on when my PCMCIA sleeve arrives...
> 
> 
>> What is the vintage of this unit?
> 
> Brand new.  S/N 4g15dw312091, on the off chance that means anything to you. :)
> Purchasing happened through another department, but I believe it was ordered
> directly from Compaq.
> 
> Interestingly, even when serial is working, I can't make it accept any
> files via xmodem.  Gives 3 retries and then aborts back to the prompt
> (at least, it's sitting at the prompt, very confused, when I abort the
> upload on the PC).  

So, how did you get RedBoot installed in the first place?

WHat are you trying to load?  It could be the case that the load command is
quitting for some other reason and your download utility (minicom, hyperterm)
just keeps trying.  RedBoot will actually take around 60 seconds to give up
on a serial download, only after some 20 retries each 3 seconds apart.

Re: [ECOS] odd serial input bug in redboot 095308 20010607 on iP

[Gary Thomas]

Can you try this experiment for me?

Cold reset the unit.  Presumably, the serial won't be working, but
don't touch it or the touch screen.  After a few seconds, the LCD 
should blank out.  Then try typing on the serial.  Does the LCD
come back on?

Another thing - do you have hardware handshaking (flow control) turned
off?  I've found this necessary, at least on my Linux machines.

RE: [ECOS] odd serial input bug in redboot 095308 20010607 on iP

[Trenton D. Adams]

I can confirm that hardware flow control has to be turned off in Windows
2000 Professional as well.

-----Original Message-----
From: ecos-discuss-owner@sources.redhat.com
[ mailto:ecos-discuss-owner@sources.redhat.com ] On Behalf Of Gary Thomas
Sent: Friday, June 08, 2001 9:47 AM
To: David Bliss
Cc: ecos-discuss@sources.redhat.com
Subject: Re: [ECOS] odd serial input bug in redboot 095308 20010607 on
iP

Can you try this experiment for me?

Cold reset the unit.  Presumably, the serial won't be working, but
don't touch it or the touch screen.  After a few seconds, the LCD 
should blank out.  Then try typing on the serial.  Does the LCD
come back on?

Another thing - do you have hardware handshaking (flow control) turned
off?  I've found this necessary, at least on my Linux machines.

Re: [ECOS] odd serial input bug in redboot 095308 20010607 on iP

[David Bliss]

On Fri, Jun 08, 2001 at 09:44:12AM -0600, Gary Thomas wrote:
>  
> Sorry, that wasn't 100% clear from your first message.

My apologies.

> > Interestingly, even when serial is working, I can't make it accept any
> > files via xmodem.  Gives 3 retries and then aborts back to the prompt
> > (at least, it's sitting at the prompt, very confused, when I abort the
> > upload on the PC).  
> 
> So, how did you get RedBoot installed in the first place?

Apparently I was unclear here also: running RedBoot via osloader from WinCE
worked just fine (for quite a period of time) with none of these diffculties.  
That's how I uploaded the current RedBoot in flash.

> WHat are you trying to load?  It could be the case that the load command is
> quitting for some other reason and your download utility (minicom, hyperterm)
> just keeps trying.  RedBoot will actually take around 60 seconds to give up
> on a serial download, only after some 20 retries each 3 seconds apart.

nanoxdemo.gz.  I haven't tried anything else, so it could be specific to this
file.

david

Re: [ECOS] odd serial input bug in redboot 095308 20010607 on iP

[David Bliss]

On Fri, Jun 08, 2001 at 09:47:05AM -0600, Gary Thomas wrote:
> 
> Can you try this experiment for me?
> 
> Cold reset the unit.  Presumably, the serial won't be working, but
> don't touch it or the touch screen.  After a few seconds, the LCD 
> should blank out.  Then try typing on the serial.  Does the LCD
> come back on?

Power off.  Power on, dock into cradle.
Comes up.  Displays 'Press Q...' on the screen.  Displayed boot banner to
serial port and screen, up to RedBoot>.  Without touching serial or screen,
wait, screen blanks.  Press enter on serial terminal a few times.  No response
on serial port, screen remains blank.  Touching the screen gives the standard
blink, then accepts touchpad input, but not serial input.

This is repeatable both with and without hardware handshaking.  Terminal
is w2k advanced server HyperTerminal.

> Another thing - do you have hardware handshaking (flow control) turned
> off?  I've found this necessary, at least on my Linux machines.

Thanks,
david

Re: [ECOS] odd serial input bug in redboot 095308 20010607 on iP

[Gary Thomas]

On 08-Jun-2001 David Bliss wrote:
> On Fri, Jun 08, 2001 at 09:44:12AM -0600, Gary Thomas wrote:
>>  
>> Sorry, that wasn't 100% clear from your first message.
> 
> My apologies.
> 
>> > Interestingly, even when serial is working, I can't make it accept any
>> > files via xmodem.  Gives 3 retries and then aborts back to the prompt
>> > (at least, it's sitting at the prompt, very confused, when I abort the
>> > upload on the PC).  
>> 
>> So, how did you get RedBoot installed in the first place?
> 
> Apparently I was unclear here also: running RedBoot via osloader from WinCE
> worked just fine (for quite a period of time) with none of these diffculties.  
> That's how I uploaded the current RedBoot in flash.
> 

Obviously this took a similar path to system initialization that got things working.
Makes sense.

>> WHat are you trying to load?  It could be the case that the load command is
>> quitting for some other reason and your download utility (minicom, hyperterm)
>> just keeps trying.  RedBoot will actually take around 60 seconds to give up
>> on a serial download, only after some 20 retries each 3 seconds apart.
> 
> nanoxdemo.gz.  I haven't tried anything else, so it could be specific to this
> file.
>

Try uncompressing the file on the server (host) end.  Then try the load command
  RedBoot> load -r -b 0x20000 -m ym
  RedBoot> go 0x20040

Re: [ECOS] odd serial input bug in redboot 095308 20010607 on iP

[David Bliss]

On Fri, Jun 08, 2001 at 10:19:45AM -0600, Gary Thomas wrote:
> 
> Obviously this took a similar path to system initialization that got things working.
> Makes sense.

That's what I thought, too.

> Try uncompressing the file on the server (host) end.  Then try the load command
>   RedBoot> load -r -b 0x20000 -m ym
>   RedBoot> go 0x20040

Hm.  It seems some overzealous web client decompressed this while downloading
it.  Without bothering to change the extension or tell me about it of course.
Naturally this confused the downloader.  It seems to work now after I
recompressed it.  However, I get horrible screen glitches while uploading --
looks like random blocks of the screen blitted to random other locations 15-20 
times per second.  (But only using load -d, not using load...)  Is this 
expected behaviour? :)

Curioser and curiouser.  If I upload the compressed version with load -d, it
complains of a CRC error.  If I upload the uncompressed version and go 0x20040,
I get '$T050f:4c000200;0d:e4b80000;#79' and '$T050f:4c000200;0d:e4b80000;#79'
in response to every keystroke (alternately).  Any  clue what *that* is?

david

Re: [ECOS] odd serial input bug in redboot 095308 20010607 on iP

[Gary Thomas]

On 08-Jun-2001 David Bliss wrote:
> On Fri, Jun 08, 2001 at 10:19:45AM -0600, Gary Thomas wrote:
> 
>> Try uncompressing the file on the server (host) end.  Then try the load command
>>   RedBoot> load -r -b 0x20000 -m ym
>>   RedBoot> go 0x20040
> 
> Hm.  It seems some overzealous web client decompressed this while downloading
> it.  Without bothering to change the extension or tell me about it of course.
> Naturally this confused the downloader.  It seems to work now after I
> recompressed it.  However, I get horrible screen glitches while uploading --
> looks like random blocks of the screen blitted to random other locations 15-20 
> times per second.  (But only using load -d, not using load...)  Is this 
> expected behaviour? :)

The screen flicker is a hardware problem.  Since RedBoot is executing from Flash,
certain patterns can saturate the memory bus which causes the LCD to flicker.

> 
> Curioser and curiouser.  If I upload the compressed version with load -d, it
> complains of a CRC error.  If I upload the uncompressed version and go 0x20040,
> I get '$T050f:4c000200;0d:e4b80000;#79' and '$T050f:4c000200;0d:e4b80000;#79'
> in response to every keystroke (alternately).  Any  clue what *that* is?

Yes, this is GDB telling you that the program dies at location 0x0024c00 which
as far as I can tell is somewhere in the NanoX initialization code.  You can see
more by disconnecting minicom/hyperterm and then connecting to the board via
GDB.
  arm-elf-gdb -nw
  (gdb) tar rem /dev/ttyS0
  (gdb) info reg

Did you also load the eCos ROM FS and install it?  It may be the case that it
was scrogged as well.

Re: [ECOS] odd serial input bug in redboot 095308 20010607 on iP

[David Bliss]

On Fri, Jun 08, 2001 at 11:03:14AM -0600, Gary Thomas wrote:
> 
> The screen flicker is a hardware problem.  Since RedBoot is executing from Flash,
> certain patterns can saturate the memory bus which causes the LCD to flicker.

Ok.

> > Curioser and curiouser.  If I upload the compressed version with load -d, it
> > complains of a CRC error.  If I upload the uncompressed version and go 0x20040,
> > I get '$T050f:4c000200;0d:e4b80000;#79' and '$T050f:4c000200;0d:e4b80000;#79'
> > in response to every keystroke (alternately).  Any  clue what *that* is?
> 
> Yes, this is GDB telling you that the program dies at location 0x0024c00 which
> as far as I can tell is somewhere in the NanoX initialization code.  You can see
> more by disconnecting minicom/hyperterm and then connecting to the board via
> GDB.
>   arm-elf-gdb -nw
>   (gdb) tar rem /dev/ttyS0
>   (gdb) info reg
> 
> Did you also load the eCos ROM FS and install it?  It may be the case that it
> was scrogged as well.

I don't have a gdb for arm/elf; I can probably install it over the weekend if
it would be useful (which it sounds like it would).  I did not load the eCos
ROMFS; the web page indicates that it's only needed if you want to burn the
nanox demo to flash.

I've just installed ecosroms.gz and reinstalled nanoxdemo; same result.
I guess I'll have to wait until I get an arm/elf debugger set up?

I do wonder why it's giving me such trouble, though.  I was under the
impression that this stuff was pretty extensively used.

Thanks,
david

Re: [ECOS] odd serial input bug in redboot 095308 20010607 on iP

[Gary Thomas]

On 08-Jun-2001 David Bliss wrote:
> I don't have a gdb for arm/elf; I can probably install it over the weekend if
> it would be useful (which it sounds like it would).  I did not load the eCos

Absolutely useful :-)

> ROMFS; the web page indicates that it's only needed if you want to burn the
> nanox demo to flash.
> 
> I've just installed ecosroms.gz and reinstalled nanoxdemo; same result.
> I guess I'll have to wait until I get an arm/elf debugger set up?
> 
> I do wonder why it's giving me such trouble, though.  I was under the
> impression that this stuff was pretty extensively used.
> 

It is and should be pretty stable.  However, some things are moving targets
and not everything keeps up to date.  The iPAQ has changed somewhat since
I wrote that demo (and the instructions) and not everything may be up to date.

That said, it did run fine for me when I tested it here today.

After you download the nanoxdemo, try running 'cksum' on the image.  Then
compare it with what you get from the host.
        RedBoot> lo -r -d -b 0x20000 nanoxdemo.gz
        Raw file loaded 0x00020000-0x00104a50
        RedBoot> cksum -b 0x20000 -l 0xe4a50
        POSIX cksum = 0x1ba7e900 (463988992)

        $ cksum /tftpboot/nanoxdemo 
        463988992 936528 /tftpboot/nanoxdemo

The numbers in () should match.

Note: I just found out why you were having trouble with the "lo -d" setup
above.  It was because of a buffer overflow in the latest version of RedBoot
on the iPAQ (which you have from June 7 - moving target, remember).  I've now
repaired this and new versions will appear sometime [hopefully soon].  In any
case, you can verify the values against the non-compressed version.

Re: [ECOS] odd serial input bug in redboot 095308 20010607 on iP

[Jonathan Larmour]

Gary Thomas wrote:
> 
> Note: I just found out why you were having trouble with the "lo -d" setup
> above.  It was because of a buffer overflow in the latest version of RedBoot
> on the iPAQ (which you have from June 7 - moving target, remember).  I've now
> repaired this and new versions will appear sometime [hopefully soon]. 

They're up now off of http://sources.redhat.com/ecos/boards/ipaq.html

Jifl
-- 
Red Hat, Rustat House, Clifton Road, Cambridge, UK. Tel: +44 (1223) 271062
Maybe this world is another planet's Hell -Aldous Huxley || Opinions==mine
Come to the Red Hat TechWorld open source conference in Brussels!
Keynotes, techie talks and exhibitions    http://www.redhat-techworld.com/